<?xml version="1.0" standalone="yes"?>
<Paper uid="H92-1057">
  <Title>Experimental Results for Baseline Speech Recognition Performance using Input Acquired from a Linear Microphone Array</Title>
  <Section position="7" start_page="288" end_page="289" type="concl">
    <SectionTitle>
6. Conclusion
</SectionTitle>
    <Paragraph position="0"> An intricate experiment has been developed to quantify the effects of alternative acoustic environments on speech-recognition systems. The performance of an HMM-based alphadigit recognizer was reduced about 12% when input was converted from high-fidelity, close-talking input to either a single remote microphone or the output of a delay-and-sum beamformer using an eight-microphone linear array under quiet conditions. Beamforming did significantly improve performance over that of a single microphone for low signal-to-noise ratios and is thus advantageous in the presence of acoustic interference.</Paragraph>
    <Paragraph position="1"> More importantly, though, the work establishes an automated procedure for reconstructing a given database in a new environment, permitting the evaluation of acoustic-input devices. Such a structured methodology has allowed the determination of baseline performance and now future improvements can be appropriately measured.</Paragraph>
  </Section>
class="xml-element"></Paper>